[SOLVED] User Action is not identified second time for same indent

Rasa Open Source
1

Hi

I am trying to write a simple bot using RASA.

I train my RASA NLU and Core with data

In RASA NLU markdown file I gave input as below

## intent:greet
- Good morning
- hello
- hi
## intent:bye
- goodbye
- goodnight

In my domain file I gave like

intents:
  - greet
  - bye

templates:
  utter_greet:
    - text: "bot : hello"
  utter_bye:
    - text: "bye was nice talking to you"
actions:
  - utter_greet
  - utter_bye

My story file

## greet
* greet
	- utter_greet
* bye
	- utter_bye

My situation is

  1. When the user input “hi” the intent is properly identified and the bot say “bot : hello”
  2. But the when user again input “hi” the intent is identified as greet and the bot action is not identified as utter_greet

From my debug log

Creating a new tracker for id 'default'.
Received user message 'hi' with intent '{'name': 'greet', 'confidence': 0.9335246682167053}' and entities '[]'
Logged UserUtterance - tracker now has 2 events
Current slot values: 

Current tracker state [None, {}, {'intent_greet': 1.0, 'prev_action_listen': 1.0}]
There is a memorised next action '3'
Predicted next action using policy_1_MemoizationPolicy
Predicted next action 'utter_greet' with prob 1.00.
Action 'utter_greet' ended with events '[]'
Bot utterance 'BotUttered(text: bot : hello, data: {
  "elements": null,
  "buttons": null,
  "attachment": null
})'
Current tracker state [{}, {'intent_greet': 1.0, 'prev_action_listen': 1.0}, {'intent_greet': 1.0, 'prev_utter_greet': 1.0}]
There is a memorised next action '0'
Predicted next action using policy_1_MemoizationPolicy
Predicted next action 'action_listen' with prob 1.00.
Action 'action_listen' ended with events '[]'
Recreating tracker for id 'default'
Received user message 'hi' with intent '{'name': 'greet', 'confidence': 0.9335246682167053}' and entities '[]'
Logged UserUtterance - tracker now has 6 events
Current slot values: 

Current tracker state [{'intent_greet': 1.0, 'prev_action_listen': 1.0}, {'intent_greet': 1.0, 'prev_utter_greet': 1.0}, {'intent_greet': 1.0, 'prev_action_listen': 1.0}]
There is no memorised next action
Predicted next action using policy_2_KerasPolicy
Predicted next action 'action_listen' with prob 0.87.
Action 'action_listen' ended with events '[]'
Recreating tracker for id 'default'
Received user message 'hi' with intent '{'name': 'greet', 'confidence': 0.9335246682167053}' and entities '[]'
Logged UserUtterance - tracker now has 8 events
Current slot values:
2

you will have to add more instances in your stories to make bot understand different scenarios for your tasks.

3

separating this into two stories should fix that:

## greet
* greet
	- utter_greet

## bye
* bye
	- utter_bye
4

Thank you … This fixed the issue.

5

@akelad I am facing the same issue…No I want same story to be repeated…Like this your input: hi # here bot is listening i.e (action_listen) bot: hello your input: hi # from here next predicted action is every time (action_listen) your input: your input: If i am going exactly as per the story and complete the story(utter_goodbye) and then again if i am starting with “hi” , every time next predicted action is action_listen.